cantillation/Teamim-medium_WeightDecay-0.005_Augmented_Old-Data_WithSRT_date-05-04-2025

he-cantillation

This model is a fine-tuned version of openai/whisper-medium on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.3512
• Wer: 37.4546
• Avg Precision Exact: 0.4711
• Avg Recall Exact: 0.4799
• Avg F1 Exact: 0.4744
• Avg Precision Letter Shift: 0.4938
• Avg Recall Letter Shift: 0.5053
• Avg F1 Letter Shift: 0.4978
• Avg Precision Word Level: 0.5066
• Avg Recall Word Level: 0.5185
• Avg F1 Word Level: 0.5107
• Avg Precision Word Shift: 0.6878
• Avg Recall Word Shift: 0.7144
• Avg F1 Word Shift: 0.6977
• Precision Median Exact: 0.4
• Recall Median Exact: 0.4353
• F1 Median Exact: 0.4211
• Precision Max Exact: 1.0
• Recall Max Exact: 1.0
• F1 Max Exact: 1.0
• Precision Min Exact: 0.0
• Recall Min Exact: 0.0
• F1 Min Exact: 0.0
• Precision Min Letter Shift: 0.0
• Recall Min Letter Shift: 0.0
• F1 Min Letter Shift: 0.0
• Precision Min Word Level: 0.0
• Recall Min Word Level: 0.0
• F1 Min Word Level: 0.0
• Precision Min Word Shift: 0.0
• Recall Min Word Shift: 0.0
• F1 Min Word Shift: 0.0

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 1e-05
• train_batch_size: 16
• eval_batch_size: 2
• seed: 42
• optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 1000
• training_steps: 20000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Avg Precision Exact	Avg Recall Exact	Avg F1 Exact	Avg Precision Letter Shift	Avg Recall Letter Shift	Avg F1 Letter Shift	Avg Precision Word Level	Avg Recall Word Level	Avg F1 Word Level	Avg Precision Word Shift	Avg Recall Word Shift	Avg F1 Word Shift	Precision Median Exact	Recall Median Exact	F1 Median Exact	Precision Max Exact	Recall Max Exact	F1 Max Exact	Precision Min Exact	Recall Min Exact	F1 Min Exact	Precision Min Letter Shift	Recall Min Letter Shift	F1 Min Letter Shift	Precision Min Word Level	Recall Min Word Level	F1 Min Word Level	Precision Min Word Shift	Recall Min Word Shift	F1 Min Word Shift
0.2285	0.1185	1000	0.9052	73.4897	0.1965	0.2129	0.2032	0.2284	0.2502	0.2371	0.2475	0.2706	0.2563	0.4222	0.4817	0.4463	0.0938	0.1127	0.1039	1.0	1.0	1.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0
0.1115	0.2370	2000	0.6479	56.3993	0.3312	0.3484	0.3382	0.3643	0.3860	0.3726	0.3819	0.4025	0.3891	0.5813	0.6327	0.6002	0.2088	0.2353	0.2193	1.0	1.0	1.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0
0.0859	0.3555	3000	0.5962	53.4888	0.3662	0.3739	0.3690	0.3988	0.4088	0.4023	0.4157	0.4260	0.4188	0.6048	0.6353	0.6160	0.2300	0.2353	0.2286	1.0	1.0	1.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0
0.0889	0.4740	4000	0.5403	50.5361	0.3794	0.3933	0.3842	0.4088	0.4254	0.4145	0.4255	0.4424	0.4310	0.6163	0.6560	0.6303	0.25	0.2778	0.2623	1.0	1.0	1.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0
0.0534	0.5925	5000	0.5052	49.8694	0.3836	0.3903	0.3853	0.4164	0.4243	0.4182	0.4330	0.4421	0.4345	0.6240	0.6507	0.6330	0.25	0.2632	0.2543	1.0	1.0	1.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0
0.0388	0.7110	6000	0.4644	47.3135	0.3987	0.4090	0.4020	0.4284	0.4409	0.4323	0.4440	0.4569	0.4476	0.6367	0.6685	0.6477	0.2857	0.3077	0.2909	1.0	1.0	1.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0
0.0485	0.8295	7000	0.4162	45.5045	0.4286	0.4373	0.4318	0.4570	0.4668	0.4602	0.4710	0.4813	0.4740	0.6585	0.6821	0.6666	0.3220	0.3333	0.3243	1.0	1.0	1.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0
0.0556	0.9480	8000	0.4246	45.8386	0.4247	0.4369	0.4298	0.4519	0.4657	0.4571	0.4682	0.4808	0.4723	0.6570	0.6876	0.6684	0.3	0.3333	0.3131	1.0	1.0	1.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0
0.0157	1.0665	9000	0.4023	43.3484	0.4395	0.4483	0.4429	0.4669	0.4766	0.4702	0.4820	0.4910	0.4848	0.6722	0.6952	0.6805	0.3333	0.3507	0.3436	1.0	1.0	1.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0
0.0295	1.1850	10000	0.4139	44.4629	0.4229	0.4358	0.4276	0.4487	0.4645	0.4542	0.4629	0.4796	0.4683	0.6422	0.6758	0.6543	0.2927	0.3171	0.3000	1.0	1.0	1.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0
0.035	1.3035	11000	0.4131	43.0376	0.4287	0.4419	0.4340	0.4505	0.4668	0.4568	0.4643	0.4813	0.4704	0.6392	0.6786	0.6538	0.2934	0.3333	0.3068	1.0	1.0	1.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0
0.0318	1.4220	12000	0.3858	41.0069	0.4437	0.4532	0.4471	0.4674	0.4793	0.4713	0.4808	0.4936	0.4849	0.6617	0.6924	0.6727	0.3263	0.3514	0.3352	1.0	1.0	1.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0
0.0225	1.5405	13000	0.3853	43.1748	0.4364	0.4438	0.4391	0.4604	0.4696	0.4636	0.4731	0.4836	0.4768	0.6536	0.6757	0.6617	0.3231	0.3333	0.3269	1.0	1.0	1.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0
0.0138	1.6590	14000	0.3712	40.2381	0.4454	0.4547	0.4487	0.4688	0.4795	0.4726	0.4822	0.4944	0.4861	0.6652	0.6926	0.6751	0.3333	0.3624	0.3478	1.0	1.0	1.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0
0.0244	1.7775	15000	0.3491	40.3840	0.4552	0.4607	0.4570	0.4789	0.4854	0.4810	0.4914	0.4990	0.4938	0.6692	0.6887	0.6760	0.35	0.3636	0.3589	1.0	1.0	1.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0
0.0327	1.8960	16000	0.3481	38.7077	0.4564	0.4673	0.4604	0.4795	0.4928	0.4843	0.4935	0.5073	0.4981	0.6744	0.7053	0.6854	0.3563	0.3767	0.3636	1.0	1.0	1.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0
0.012	2.0145	17000	0.3537	38.2292	0.4712	0.4797	0.4745	0.4942	0.5046	0.4979	0.5071	0.5173	0.5103	0.6878	0.7134	0.6973	0.4049	0.4231	0.4079	1.0	1.0	1.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0
0.0048	2.1330	18000	0.3589	38.2672	0.4642	0.4732	0.4677	0.4870	0.4989	0.4912	0.4992	0.5110	0.5029	0.6806	0.7064	0.6900	0.3835	0.4128	0.3883	1.0	1.0	1.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0
0.0067	2.2515	19000	0.3534	37.7172	0.4712	0.4786	0.4740	0.4944	0.5045	0.4979	0.5070	0.5170	0.5101	0.6912	0.7149	0.6997	0.4	0.4326	0.4194	1.0	1.0	1.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0
0.007	2.3699	20000	0.3512	37.4546	0.4711	0.4799	0.4744	0.4938	0.5053	0.4978	0.5066	0.5185	0.5107	0.6878	0.7144	0.6977	0.4	0.4353	0.4211	1.0	1.0	1.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0

Framework versions

• Transformers 4.49.0
• Pytorch 2.6.0+cu126
• Datasets 2.12.0
• Tokenizers 0.20.1